Improved Observation Operators for Near Surface Wind and Temperature Assimilation in the RTMA

NOAA’s Real Time and UnRestricted Mesoscale Analysis systems (RTMA and URMA respectively) produce ~2.5 km near-surface analyses of sensible weather elements at hourly intervals for use in nowcasting, calibration, and bias correction - notably in NOAA’s National Blend of Models. This effort involves the assimilation of a variety of observations, including those from non-standard, crowd-sourced mesonet networks such as CWOP.

Ongoing validation efforts have shown, however, that the assimilation of these non-standard observations can lead to a low wind speed bias in the analysis of 10m winds. While some anemometers may be sited in regions with poor exposure, this is also in part due to a lack of available metadata providing sensor heights. In the current analysis algorithm, it is assumed these observations are taken from 10 m AGL, however it is more likely that the observations are much lower to the ground - thus introducing a slow bias in the resulting analysis when these winds are assimilated. In some cases sensor height information is readily available and the information can be used effectively to enhance the observation operator in the analysis algorithm, yielding an improved analysis. While part of this work involves a substantial metadata gathering effort, it also involves introducing and leveraging the appropriate capability in the analysis scheme. The latter will be discussed in this presentation.

Further work is also underway to address problems in the analysis of temperature across sharp gradients in terrain (as has been noted by forecasters). In such cases the discrepancy between the resolved terrain in the 2.5 km analysis and the station elevation, which may often differ considerably, can introduce unphysical temperature gradients when this difference in elevation is not taken into consideration.

This presentation will discuss ongoing efforts toward improving the assimilation of near-surface wind and temperature observations through the enhancement of their respective observation operators.